How Av Works in Practice

Although for national parliaments the AV system is only used in Australia, Fiji and Papua New Guinea (recently adopted in the last two), it is being tried for some local elections in America.

In the UK we already use AV for elections for mayors, and the system has proven to work well where there are two clear front runners. The suitability of the system for general elections where we are electing a government, and where there may be a spread of candidates and multiple constituencies that are three way marginals is not so clear.

In a recent election in North Carolina there was a wide field [SEE BALLOT PAPER] and the two front runner candidates only had 35% of the first preference votes between them: Chaos ensued.

“I think we’ve seen the last of IRV voting in North Carolina,” State Board of Elections Executive Director Gary Bartlett told me this week. [IRV = AV]

I hope so. And I’m going to beat the dying horse a little more just for good measure.

IRV — Instant Runoff Voting — was used in a special election for a Court of Appeals seat in November. You might remember there was a field of 13 candidates and you voted for your first, second and third choices.

Bartlett drew his opinion not because elections officials mishandled this special election and its complex, confusing procedures but because it was unpopular. Leading legislators and the governor didn’t like it, and there’s a good chance they’ll drop the flirtation with this odd brand of voting.

The outcome sure was strange.

The result on election night was that Cressie Thigpen led with 395.220 first-choice votes, or 20 percent of the total.

Doug McCullough was second with 295,619 votes, or 15 percent.

Because Thigpen didn’t win a majority, he and McCullough advanced to the “instant runoff.” This was neither instant nor a runoff. Rather, it meant that second- and third-choice votes would be added to their totals to decide the ultimate winner.

About seven weeks later, that was determined to be McCullough with 543,980 votes, edging Thigpen, who had 537,325.

McCullough had turned a deficit of nearly 100,000 votes into a winning margin of 6,655 votes.

But here’s a funny thing. Bartlett sent me raw vote totals last week: the numbers of 1, 2 and 3 votes for all 13 candidates in all 100 counties. When I got around to adding them up, I found that Thigpen still had the most votes by nearly 100,000:

718,042 for Thigpen

618,431 for McCullough.

How was that?

The raw vote totals include about a quarter-million votes that were tossed out in the second round of counting.

Some of those were multiple votes for one candidate. For example, if anyone used his first, second and third votes for Thigpen, all three votes are included in the raw total but only one would count in the official tally.

Also, voters whose first choice was Thigpen or McCullough had their ballots discarded at that point. For example, if someone voted for Thigpen as first choice and McCullough as second choice, the vote for McCullough was not counted in the second round of counting. The idea was that this voter should not have his second-choice vote cancel his first-choice vote.

So, for one reason or the other, Thigpen lost 180,717 votes; McCullough lost only 74,451 votes — a huge difference that statisticians might suggest was improbable.

Thigpen must have had more people who voted for him multiple times than did McCullough. And more of McCullough’s first-choice voters must have given their second- or third-choice votes to Thigpen than the other way around. Either way, the math worked out very badly for Thigpen.

Analyzing exactly what happened — an exercise the legislature’s program evaluation division should undertake — would require examining all ballots and finding out exactly how people voted.

My conclusion is it was convoluted, delivering a result that is hard to understand. It will be better to put this horse out to pasture.

18 comments on “How Av Works in Practice

  1. it took seven weeks to get a result, and there was voter disatisfaction in the result not being as they expected.

    Ok, for a North Carolina judge perhaps that is acceptable, but for electing governments?

  2. Harsanyi_Janos on said:

    I had missed the seven week delay. That is clearly unacceptable, but machine counting would surely make that unnecessary?

  3. Lets not forget that in the states the existing parties have vested interests in resisting electoral reform. LOok at what happened in the ’50s, New York used STV to elect it’s council, but that6 got reversed when some republicans got elected 🙂

    I note the objections in the article are based on the fact that, contrary to some claims about AV, it does *not* give multiple votes, the vote rests with the highest ranked candidate.

    Why there was a seven week delay (presumably legal challenges) doesn’t stem from the electoral system, and the critciism here sound like they’re from someone who (deliberately) doesn’t understand the methodology, and is seekign to complicate a quite simple situation.

  4. Also, it’s clear that the above ballot was not using full AV, but a limited version, if they had allowed a preference to be expressed for all candidates (as in Australia) the count may have been better handled, and less confusing.

  5. Michael C on said:

    Yes this is a variant of the Supplementary Vote (used for the Mayorals) not AV proper. No idea why it took 7 weeks but the inference that our GE’s would look anything like this under AV is a total red herring.

  6. This is not how the proposed AV system works because while the Mayoral uses this system where you have an instant run off between the top two candidates AV does not and excludes candidates from the bottom until one candidate has more than 50% of the remianing votes still in play. ie the candidate placed third in the first round *can* win, unlike in the Mayorals.

  7. BUt when translated into electing a legislature of multiple members with single representatives for geographic constutuencies we see that the party that gets not only the most first preference votes, but also more than 50% of subsequent prefernces can still lose the election, as has happened in Australia FIVE TIMES since 1946!

    Why do no supporters of AV address this?

    It seems that AV is much more sensitive to boundary gerrymandering than FPTP.

  8. flying train on said:

    total red herring article

    irv can in no way be described as av and this is NOT how it works in practice.

    in av, you simply knockout the candidate with the least votes and re-distribute his/her 2nd prefs and so on until someone has 50%. far easier to count and happens a lot lot quicker. no problem of votes cancelling out another as 1st candidate is out already.

    irv is totally pash, suits a two clear candidates yes but that is not bitain and is in no way suitable.

    crap article, sorry

  9. flying train on said:


    exactly why av is not really the best compromise. stv with multi-member is far better than anything else suitable and will be the way governments are elected in the not to distant future.

    scotland and i dare say wales will soon change the way it elects its devolved parliaments, scot local gov is already there.

  10. #10

    “scotland and i dare say wales will soon change the way it elects its devolved parliaments”

    really? why would they do that?

    The system they use now is perfectly serviceabe, and is used in Germany, New Zealand, and elsewhere.

  11. flying train on said:


    because living in scotland and being around political people i get a real feel that ams is not

    1 particularly liked
    2 suitable
    3 having much of a future.


    a) a lot of people i know don’t really like the idea of constituency and list msps. there is friction between the two, both between and within parties. this would not go to a referendum, so it does not matter what the public believe or not about something but about how the politicians think and i don’t see any love for ams. we would not go to fptp or av, so any potential change would be towards stv. you could also combine 2 constituency seats with 1 or 2 list seats to make a mix of 3 or 4 (or add one for 4 or 5) member seats without the areas being huge (always muted as a problem for westminister elections)

    b) we have at the last count 4 (i think but happy to be corrected) different electoral systems on the go in the scotland, fptp, ams, stv and party list for euros. in terms of how you count this and cost, it is not particularly suitable to have so many different systems. it costs councils a lot of money to have 4 different systems. To standardise, and thus have the same computers for a couple of different elections makes a lot of sense

    c) the idea of a an msp beaten in a constituency but returned on the list though up to now not great, has not caused problems. what happens when a deeply unpopular / potentially corrupt / incompetent msp is destroyed is his / her constituency but is retained on the list? ams would not outlast this.

    wales i maybe do not know about but the same as above is true apart i think from stv at local gov. all this would require is either the nationalists or the libs to make it a condition of coalition and it would happen.

  12. christian h. on said:

    While I actually agree with voting “no” on AV because it is a distraction from real PR, Andy’s technical points are quite confused (no surprise since he didn’t understand London mayoral IRV at the time). Before I come to why it is confused, I’d say that ironically the fact that Andy who is fairly educated (not a dig Andy) and smart has trouble understanding the system’s implications might itself seem like a counter-argument – on that below.

    1. I has been pointed out already that AV is not the same as IRV. It should also be made clear that had NC used a more common non-IRV “open primary” system, the top two candidates – each of whom was very far from a majority – would have gone to a run-off. The total time to election would have been about the same. Finally, if FPTP had been used in this election, a candidate with the support of 15% of the voters would have been elected. In other words, the problem here was not IRV but the open primary leading to a very large number of viable candidates. No single member constituency voting technique can deal with that well, although true AV works best.

    2. It is of course possible to come up with cases in which the result of an AV vote seemingly is against the voters’ wishes. this is a result of a mathematical theorem which says that NO voting system can be designed that avoids this. FPTP constantly leads to such distortions although they are not so obvious to see since there are no lower preference votes to inspect.

    3. (True) AV is not completely immune to “tactical voting” but almost so (again I could dig up the research if someone is interested). It is not necessary for the voter to understand the mechanics of the process (any more than it is necessary to know how a voting machine works to use it); simply rank candidates in order of preference.

  13. Strategist on said:

    #14 “I actually agree with voting “no” on AV because it is a distraction from real PR”

    Christian, this is a colossal error to make. Your “no” vote will be kindly interpreted for you by David Cameron, John Reid, and fucking David Blunkett as a vote in favour of First Past the Post and a nice cosy Tory/Labour duopoly, and any chance of PR will be buried under concrete for a generation.
    Check out Jim Jepps’ post on the 1979 Scottish referendum, where those who voted no to devolution because it was “a distraction from independence” kindly had their no vote interpreted for them by the Tories as a green light to ruthlessly fuck the country over from Westminster for 18 years.

    Don’t be Cameron’s useful idiot!

  14. Red Bandits on said:

    Just confirms my view that AV is the bone the Tories threw the yellow poodles
    We should have nothing to do with it, it aint useful except to shore up the yellow poodles
    Abstain or Vote No Stuff the Yellow Poodles

  15. David Ellis on said:

    #15 Don’t be Nick Griffin’s useful idiot. By campaigning hard in two core seats and just putting up candidates in many more to pick up a couple of thousand votes here and there the television, internet and PR would hand the BNP at least 20-25 seats without most of them having to leave the comfort of their bedrooms. Constituency democracy is the most democratic form of government under capitalism (where which ever system is biased toward the rich and their money) because it recognizes geographical differences and particularities. AV and PR shit all over that. The two-party system is a luxury which the ruling class can no longer afford and we should let them do the heavy lifting when it comes to smashing it up. We on the other hand only need to put forward a socialist programme and seek to mobilise millions in strikes, demos, sit-ins, factory and workplace and community committees against the cuts. In the meantime we can give the Coalition a bloody political nose by pissing all over the Lib Dem’s search for legitimacy and a mandate: No2AV Yes to Socialism.

    As for the Scottish and Welsh issues, I for one will be voting No 2 AV and Ie Dros Gymru.

  16. christian h. on said:

    Strategist: first don’t worry, I will not have a vote. On the issue itself, the question is whether AV is a step in the right direction. I’d suggest the Australian experiences shows it really isn’t. There is barely more representation of minor parties in their parliament than in the UK. It seems to me if you wanted some real progress without going all the way to PR (ignore Ellis he’s full of it as usual) you should go with multi-member constituencies (as for the Euro elections).